Cloud Foundry Day at KubeCon + CloudNativeCon 2018

There is no need to expain the importance and value of log aggregation machanisms designed for provisioning complex and large-scale environments. Many companies with enterprise-ready implementations of Cloud Foundry have sussessful experience with the ELK stack as an application and platform log management stack. In the meantime, same companies are looking into or already using Kubernetes as an additional option for application runtime alongside with CF. This trend heavily requires services to support different types of envronments and architectures. During this talk, we will tell how to benefit from using Logsearch for Cloud Foundry as a single and secure instrument for logging hybrid environments. 

Cloud Foundry provides heavy lifting of running stateless applications in the cloud. But what about data services? The default method is to create a BOSH release for every service. In this presentation, we will show you an alternative approach, when a data service with its service broker is deployed to Kubernetes. First, we will deploy a MariaDB instance using Helm on top of a Kubernetes cluster running on the Google Cloud Platform. Then, we will put the MariaDB service broker into a Docker container, deploy it to Kubernetes as a replicated pod, and expose to Cloud Foundry using the Service Load Balancer. Finally, we will register the service broker in the Cloud Foundry marketplace, provision a service instance, and connect our stateless application running on Cloud Foundry to this instance.

Putting CF on Kubernetes is easy (just use a bosh CPI, or use fissile to convert bosh releases to containers and then deploy that). But that approach doesn't really give you any of the benefits of Kubernetes. You still have to manage two schedulers, you don't get to use any of the nice Kubernetes tools to interact with your app containers (because they're inside Diego, not Kubernetes) and your operators need to know how to debug and maintain two different systems, precariously balanced on top of each other.

Cube is a proof of concept for combining CF and Kubernetes the right way: by using Kubernetes as a (pluggable) scheduler for CF, alongside Diego. CF provides the developer abstraction, letting developers focus on stateless apps and rapid development. Under the covers Cube converts apps in to first-class Kubernetes Deployments and keeps them in sync: The best of both worlds.